model_name: TACE-OAM-L model_key: tace-oam-l model_version: '0.2.0' date_added: '2026-04-09' date_published: '2026-04-09' authors: - name: Zemin Xu affiliation: ShanghaiTech University, Nanjing University email: xv_chana@foxmail.com github: https://github.com/xvzemin corresponding: false - name: Wenbo Xie affiliation: ShanghaiTech University email: xiewb1@shanghaitech.edu.cn orcid: https://orcid.org/0000-0001-5321-2494 corresponding: true - name: P. Hu affiliation: ShanghaiTech University email: hupj@shanghaitech.edu.cn orcid: https://orcid.org/0000-0002-6318-1051 corresponding: true trained_by: - name: Zemin Xu affiliation: ShanghaiTech University, Nanjing University email: xv_chana@foxmail.com github: https://github.com/xvzemin corresponding: false repo: https://github.com/xvzemin/tace checkpoint_url: https://huggingface.co/xvzemin/tace-foundations/resolve/main/TACE-OAM-L.pt url: https://github.com/xvzemin/tace pr_url: https://github.com/janosh/matbench-discovery/pull/321 doi: https://doi.org/10.57967/hf/7458 paper: https://arxiv.org/abs/2509.14961 openness: OSOD # see `Open` enum in matbench_discovery/enums.py train_task: S2EFS # see `Task` enum in matbench_discovery/enums.py test_task: IS2RE-SR # see `Task` enum in matbench_discovery/enums.py targets: EFS_G # see `Targets` enum in matbench_discovery/enums.py model_type: UIP # see `ModelType` enum in matbench_discovery/enums.py model_params: 82_899_905 trained_for_benchmark: false n_estimators: 1 license: code: MIT code_url: https://github.com/xvzemin/tace/blob/main/LICENSE.md checkpoint: CC-BY-4.0 # URL that points to the license file for the model checkpoint, not the checkpoint file itself. checkpoint_url: https://creativecommons.org/licenses/by/4.0/legalcode hyperparams: # strongly recommended to list relaxation hyperparams max_force: 0.02 max_steps: 500 ase_optimizer: FIRE cell_filter: FrechetCellFilter optimizer: AdamW graph_construction_radius: 6.0 max_neighbors: .inf epochs: 3 batch: 1024 loss_weights: energy: 1 force: 1 stress: 0.1 huber_delta: [0.015, 0.040, 0.1] initial_learning_rate: 0.008 num_channel: 64 lmax: 5 l1l2: <= # when tp, only l1 < l2 are allowed num_layers: 5 correlation: 2 training_cost: # list any hardware used to train the model and for how long Nvidia H20 GPUs: {amount: 8, hours: 105, cost: 840} requirements: # strongly recommended python: '3.12.11' torch: '2.9.1' torch-geometric: '2.7.0' pytorch-lightning: '2.5.5' tace: '>=0.2.0' # recommended newest github commit training_set: [OMat24, sAlex, MPtrj] notes: Description: | Support both Spherical tensor and irreducible Cartesian tensor. Steps: | Training performed by: (1) pre-training on 3 epoch OMat24 (lr-8e-3, loss-1:1:0.1); (2) fine-tuning on on 3 epoch MPtrj+sAlex (lr-1e-4, loss-1:1:0.1). metrics: phonons: kappa_103: pred_file: models/tace/tace-oam-l/2026-04-09-kappa-103-FIRE-dist=0.03-fmax=0.0001-symprec=1e-05.json.gz pred_file_url: https://figshare.com/files/63590373 κ_SRME: 0.126 κ_SRE: 0.0502 geo_opt: pred_file: models/tace/tace-oam-l/2026-04-09-wbm-IS2RE-FIRE.jsonl.gz pred_file_url: https://figshare.com/files/63594915 struct_col: tace_structure symprec=1e-2: rmsd: 0.0606 # unitless n_sym_ops_mae: 1.7042 # unitless symmetry_decrease: 0.0479 # fraction symmetry_match: 0.8188 # fraction symmetry_increase: 0.126 # fraction n_structures: 256963 # count analysis_file: models/tace/tace-oam-l/2026-04-09-wbm-IS2RE-FIRE-symprec=1e-2-moyo=0.7.9.csv.gz analysis_file_url: https://figshare.com/files/63594918 symprec=1e-5: rmsd: 0.0606 # unitless n_sym_ops_mae: 2.1326 # unitless symmetry_decrease: 0.0603 # fraction symmetry_match: 0.6922 # fraction symmetry_increase: 0.2415 # fraction n_structures: 256963 # count analysis_file: models/tace/tace-oam-l/2026-04-09-wbm-IS2RE-FIRE-symprec=1e-5-moyo=0.7.9.csv.gz analysis_file_url: https://figshare.com/files/63594921 discovery: pred_file: models/tace/tace-oam-l/2026-04-09-wbm-IS2RE.csv.gz # tace-filtered_preds.csv.gz tace-preds.csv.gz pred_file_url: https://figshare.com/files/63590511 pred_col: e_form_per_atom_tace full_test_set: F1: 0.892 # fraction DAF: 5.127 # dimensionless Precision: 0.88 # fraction Recall: 0.904 # fraction Accuracy: 0.962 # fraction TPR: 0.904 # fraction FPR: 0.026 # fraction TNR: 0.974 # fraction FNR: 0.096 # fraction TP: 39872.0 # count FP: 5454.0 # count TN: 207417.0 # count FN: 4220.0 # count MAE: 0.02 # eV/atom RMSE: 0.067 # eV/atom R2: 0.862 # dimensionless missing_preds: 3 # count most_stable_10k: F1: 0.985 # fraction DAF: 6.347 # dimensionless Precision: 0.97 # fraction Recall: 1.0 # fraction Accuracy: 0.97 # fraction TPR: 1.0 # fraction FPR: 1.0 # fraction TNR: 0.0 # fraction FNR: 0.0 # fraction TP: 9702.0 # count FP: 298.0 # count TN: 0.0 # count FN: 0.0 # count MAE: 0.019 # eV/atom RMSE: 0.079 # eV/atom R2: 0.867 # dimensionless missing_preds: 0 # count unique_prototypes: F1: 0.91 # fraction DAF: 5.898 # dimensionless Precision: 0.902 # fraction Recall: 0.919 # fraction Accuracy: 0.972 # fraction TPR: 0.919 # fraction FPR: 0.018 # fraction TNR: 0.982 # fraction FNR: 0.081 # fraction TP: 30661.0 # count FP: 3344.0 # count TN: 178770.0 # count FN: 2713.0 # count MAE: 0.02 # eV/atom RMSE: 0.067 # eV/atom R2: 0.868 # dimensionless missing_preds: 1 # count